VISÃO COMPUTACIONAL E INTELIGÊNCIA ARTIFICIAL PARA MEDIÇÃO DO MESOCARPO DE TOMATES

Abstract

Advances in computer vision and artificial intelligence have driven innovation in agriculture, particularly in automated product inspection. In the context of Agro 4.0, standardized systems are crucial for optimizing processes and ensuring objective analyses compared to conventional methods. This study investigates the application of Computer Vision and Artificial Intelligence in measuring the mesocarp thickness of tomatoes (Solanum lycopersicum L.), an essential parameter for industrial and commercial standardization. To this end, a low-cost integrated hardware and software system was developed, based on digital image capture and processing. Statistical validation compared automated methods with physical reference measurements. The results demonstrated that the Artificial Intelligence (AI) model outperformed the conventional Computer Vision (CV) technique, achieving an accuracy of 90.76% (compared to 86.88% for CV) and a Mean Absolute Error (MAE) of 0.608 mm. The system achieved an optimized inference time of approximately 6 seconds per sample. It is concluded that the automation tool demonstrates technical feasibility and high reliability, offering promising prospects for realtime classification, selection of new varieties, and the improvement of phenotypic evaluation of fruits in vegetable farming and industry.